blob: 4411ada4a91cc7848cf3d9953b2181b6ca350917 [file] [log] [blame]
Bjorn Helgaase1e86ee2018-01-26 14:12:23 -06001// SPDX-License-Identifier: GPL-2.0
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08002/*
Bjorn Helgaasdf62ab52018-03-09 16:36:33 -06003 * Implement the AER root port service driver. The driver registers an IRQ
4 * handler. When a root port triggers an AER interrupt, the IRQ handler
5 * collects root port status and schedules work.
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08006 *
7 * Copyright (C) 2006 Intel Corp.
8 * Tom Long Nguyen (tom.l.nguyen@intel.com)
9 * Zhang Yanmin (yanmin.zhang@intel.com)
Bjorn Helgaas41cbc9e2018-06-08 08:40:00 -050010 *
11 * (C) Copyright 2009 Hewlett-Packard Development Company, L.P.
12 * Andrew Patterson <andrew.patterson@hp.com>
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +080013 */
14
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -050015#include <linux/cper.h>
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +080016#include <linux/pci.h>
Rafael J. Wysocki415e12b2011-01-07 00:55:09 +010017#include <linux/pci-acpi.h>
Alexey Dobriyand43c36d2009-10-07 17:09:06 +040018#include <linux/sched.h>
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +080019#include <linux/kernel.h>
20#include <linux/errno.h>
21#include <linux/pm.h>
22#include <linux/init.h>
23#include <linux/interrupt.h>
24#include <linux/delay.h>
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -050025#include <linux/kfifo.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090026#include <linux/slab.h>
Bjorn Helgaas256a4592018-06-08 08:39:45 -050027#include <acpi/apei.h>
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -050028#include <ras/ras_event.h>
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +080029
Bjorn Helgaas4696b822018-06-08 08:48:47 -050030#include "../pci.h"
31#include "portdrv.h"
Bjorn Helgaas23e672b2018-06-08 08:41:28 -050032
33#define AER_ERROR_SOURCES_MAX 100
Bjorn Helgaas23e672b2018-06-08 08:41:28 -050034
Rajat Jaindb89ccb2018-06-30 15:07:17 -050035#define AER_MAX_TYPEOF_COR_ERRS 16 /* as per PCI_ERR_COR_STATUS */
36#define AER_MAX_TYPEOF_UNCOR_ERRS 26 /* as per PCI_ERR_UNCOR_STATUS*/
37
Bjorn Helgaas23e672b2018-06-08 08:41:28 -050038struct aer_err_source {
39 unsigned int status;
40 unsigned int id;
41};
42
43struct aer_rpc {
44 struct pci_dev *rpd; /* Root Port device */
45 struct work_struct dpc_handler;
46 struct aer_err_source e_sources[AER_ERROR_SOURCES_MAX];
47 struct aer_err_info e_info;
48 unsigned short prod_idx; /* Error Producer Index */
49 unsigned short cons_idx; /* Error Consumer Index */
50 int isr;
51 spinlock_t e_lock; /*
52 * Lock access to Error Status/ID Regs
53 * and error producer/consumer index
54 */
55 struct mutex rpc_mutex; /*
56 * only one thread could do
57 * recovery on the same
58 * root port hierarchy
59 */
60};
61
Rajat Jaindb89ccb2018-06-30 15:07:17 -050062/* AER stats for the device */
63struct aer_stats {
64
65 /*
66 * Fields for all AER capable devices. They indicate the errors
67 * "as seen by this device". Note that this may mean that if an
68 * end point is causing problems, the AER counters may increment
69 * at its link partner (e.g. root port) because the errors will be
70 * "seen" by the link partner and not the the problematic end point
71 * itself (which may report all counters as 0 as it never saw any
72 * problems).
73 */
74 /* Counters for different type of correctable errors */
75 u64 dev_cor_errs[AER_MAX_TYPEOF_COR_ERRS];
76 /* Counters for different type of fatal uncorrectable errors */
77 u64 dev_fatal_errs[AER_MAX_TYPEOF_UNCOR_ERRS];
78 /* Counters for different type of nonfatal uncorrectable errors */
79 u64 dev_nonfatal_errs[AER_MAX_TYPEOF_UNCOR_ERRS];
80 /* Total number of ERR_COR sent by this device */
81 u64 dev_total_cor_errs;
82 /* Total number of ERR_FATAL sent by this device */
83 u64 dev_total_fatal_errs;
84 /* Total number of ERR_NONFATAL sent by this device */
85 u64 dev_total_nonfatal_errs;
86
87 /*
88 * Fields for Root ports & root complex event collectors only, these
89 * indicate the total number of ERR_COR, ERR_FATAL, and ERR_NONFATAL
90 * messages received by the root port / event collector, INCLUDING the
91 * ones that are generated internally (by the rootport itself)
92 */
93 u64 rootport_total_cor_errs;
94 u64 rootport_total_fatal_errs;
95 u64 rootport_total_nonfatal_errs;
96};
97
Bjorn Helgaas23e672b2018-06-08 08:41:28 -050098#define AER_LOG_TLP_MASKS (PCI_ERR_UNC_POISON_TLP| \
99 PCI_ERR_UNC_ECRC| \
100 PCI_ERR_UNC_UNSUP| \
101 PCI_ERR_UNC_COMP_ABORT| \
102 PCI_ERR_UNC_UNX_COMP| \
103 PCI_ERR_UNC_MALF_TLP)
104
105#define SYSTEM_ERROR_INTR_ON_MESG_MASK (PCI_EXP_RTCTL_SECEE| \
106 PCI_EXP_RTCTL_SENFEE| \
107 PCI_EXP_RTCTL_SEFEE)
108#define ROOT_PORT_INTR_ON_MESG_MASK (PCI_ERR_ROOT_CMD_COR_EN| \
109 PCI_ERR_ROOT_CMD_NONFATAL_EN| \
110 PCI_ERR_ROOT_CMD_FATAL_EN)
111#define ERR_COR_ID(d) (d & 0xffff)
112#define ERR_UNCOR_ID(d) (d >> 16)
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +0800113
Randy Dunlap7f785762007-10-05 13:17:58 -0700114static int pcie_aer_disable;
115
116void pci_no_aer(void)
117{
Bjorn Helgaas7ece1412016-09-06 16:24:37 -0500118 pcie_aer_disable = 1;
Randy Dunlap7f785762007-10-05 13:17:58 -0700119}
120
Rafael J. Wysockif1a7bfa2010-08-21 01:50:52 +0200121bool pci_aer_available(void)
122{
123 return !pcie_aer_disable && pci_msi_enabled();
124}
125
Bjorn Helgaas41cbc9e2018-06-08 08:40:00 -0500126#ifdef CONFIG_PCIE_ECRC
127
128#define ECRC_POLICY_DEFAULT 0 /* ECRC set by BIOS */
129#define ECRC_POLICY_OFF 1 /* ECRC off for performance */
130#define ECRC_POLICY_ON 2 /* ECRC on for data integrity */
131
132static int ecrc_policy = ECRC_POLICY_DEFAULT;
133
134static const char *ecrc_policy_str[] = {
135 [ECRC_POLICY_DEFAULT] = "bios",
136 [ECRC_POLICY_OFF] = "off",
137 [ECRC_POLICY_ON] = "on"
138};
139
140/**
141 * enable_ercr_checking - enable PCIe ECRC checking for a device
142 * @dev: the PCI device
143 *
144 * Returns 0 on success, or negative on failure.
145 */
146static int enable_ecrc_checking(struct pci_dev *dev)
147{
148 int pos;
149 u32 reg32;
150
151 if (!pci_is_pcie(dev))
152 return -ENODEV;
153
154 pos = dev->aer_cap;
155 if (!pos)
156 return -ENODEV;
157
158 pci_read_config_dword(dev, pos + PCI_ERR_CAP, &reg32);
159 if (reg32 & PCI_ERR_CAP_ECRC_GENC)
160 reg32 |= PCI_ERR_CAP_ECRC_GENE;
161 if (reg32 & PCI_ERR_CAP_ECRC_CHKC)
162 reg32 |= PCI_ERR_CAP_ECRC_CHKE;
163 pci_write_config_dword(dev, pos + PCI_ERR_CAP, reg32);
164
165 return 0;
166}
167
168/**
169 * disable_ercr_checking - disables PCIe ECRC checking for a device
170 * @dev: the PCI device
171 *
172 * Returns 0 on success, or negative on failure.
173 */
174static int disable_ecrc_checking(struct pci_dev *dev)
175{
176 int pos;
177 u32 reg32;
178
179 if (!pci_is_pcie(dev))
180 return -ENODEV;
181
182 pos = dev->aer_cap;
183 if (!pos)
184 return -ENODEV;
185
186 pci_read_config_dword(dev, pos + PCI_ERR_CAP, &reg32);
187 reg32 &= ~(PCI_ERR_CAP_ECRC_GENE | PCI_ERR_CAP_ECRC_CHKE);
188 pci_write_config_dword(dev, pos + PCI_ERR_CAP, reg32);
189
190 return 0;
191}
192
193/**
194 * pcie_set_ecrc_checking - set/unset PCIe ECRC checking for a device based on global policy
195 * @dev: the PCI device
196 */
197void pcie_set_ecrc_checking(struct pci_dev *dev)
198{
199 switch (ecrc_policy) {
200 case ECRC_POLICY_DEFAULT:
201 return;
202 case ECRC_POLICY_OFF:
203 disable_ecrc_checking(dev);
204 break;
205 case ECRC_POLICY_ON:
206 enable_ecrc_checking(dev);
207 break;
208 default:
209 return;
210 }
211}
212
213/**
214 * pcie_ecrc_get_policy - parse kernel command-line ecrc option
215 */
216void pcie_ecrc_get_policy(char *str)
217{
218 int i;
219
220 for (i = 0; i < ARRAY_SIZE(ecrc_policy_str); i++)
221 if (!strncmp(str, ecrc_policy_str[i],
222 strlen(ecrc_policy_str[i])))
223 break;
224 if (i >= ARRAY_SIZE(ecrc_policy_str))
225 return;
226
227 ecrc_policy = i;
228}
229#endif /* CONFIG_PCIE_ECRC */
230
Bjorn Helgaas256a4592018-06-08 08:39:45 -0500231#ifdef CONFIG_ACPI_APEI
232static inline int hest_match_pci(struct acpi_hest_aer_common *p,
233 struct pci_dev *pci)
234{
235 return ACPI_HEST_SEGMENT(p->bus) == pci_domain_nr(pci->bus) &&
236 ACPI_HEST_BUS(p->bus) == pci->bus->number &&
237 p->device == PCI_SLOT(pci->devfn) &&
238 p->function == PCI_FUNC(pci->devfn);
239}
240
241static inline bool hest_match_type(struct acpi_hest_header *hest_hdr,
242 struct pci_dev *dev)
243{
244 u16 hest_type = hest_hdr->type;
245 u8 pcie_type = pci_pcie_type(dev);
246
247 if ((hest_type == ACPI_HEST_TYPE_AER_ROOT_PORT &&
248 pcie_type == PCI_EXP_TYPE_ROOT_PORT) ||
249 (hest_type == ACPI_HEST_TYPE_AER_ENDPOINT &&
250 pcie_type == PCI_EXP_TYPE_ENDPOINT) ||
251 (hest_type == ACPI_HEST_TYPE_AER_BRIDGE &&
252 (dev->class >> 16) == PCI_BASE_CLASS_BRIDGE))
253 return true;
254 return false;
255}
256
257struct aer_hest_parse_info {
258 struct pci_dev *pci_dev;
259 int firmware_first;
260};
261
262static int hest_source_is_pcie_aer(struct acpi_hest_header *hest_hdr)
263{
264 if (hest_hdr->type == ACPI_HEST_TYPE_AER_ROOT_PORT ||
265 hest_hdr->type == ACPI_HEST_TYPE_AER_ENDPOINT ||
266 hest_hdr->type == ACPI_HEST_TYPE_AER_BRIDGE)
267 return 1;
268 return 0;
269}
270
271static int aer_hest_parse(struct acpi_hest_header *hest_hdr, void *data)
272{
273 struct aer_hest_parse_info *info = data;
274 struct acpi_hest_aer_common *p;
275 int ff;
276
277 if (!hest_source_is_pcie_aer(hest_hdr))
278 return 0;
279
280 p = (struct acpi_hest_aer_common *)(hest_hdr + 1);
281 ff = !!(p->flags & ACPI_HEST_FIRMWARE_FIRST);
282
283 /*
284 * If no specific device is supplied, determine whether
285 * FIRMWARE_FIRST is set for *any* PCIe device.
286 */
287 if (!info->pci_dev) {
288 info->firmware_first |= ff;
289 return 0;
290 }
291
292 /* Otherwise, check the specific device */
293 if (p->flags & ACPI_HEST_GLOBAL) {
294 if (hest_match_type(hest_hdr, info->pci_dev))
295 info->firmware_first = ff;
296 } else
297 if (hest_match_pci(p, info->pci_dev))
298 info->firmware_first = ff;
299
300 return 0;
301}
302
303static void aer_set_firmware_first(struct pci_dev *pci_dev)
304{
305 int rc;
306 struct aer_hest_parse_info info = {
307 .pci_dev = pci_dev,
308 .firmware_first = 0,
309 };
310
311 rc = apei_hest_parse(aer_hest_parse, &info);
312
313 if (rc)
314 pci_dev->__aer_firmware_first = 0;
315 else
316 pci_dev->__aer_firmware_first = info.firmware_first;
317 pci_dev->__aer_firmware_first_valid = 1;
318}
319
320int pcie_aer_get_firmware_first(struct pci_dev *dev)
321{
322 if (!pci_is_pcie(dev))
323 return 0;
324
Alexandru Gagniuc7af02fc2018-07-03 18:27:43 -0500325 if (pcie_ports_native)
326 return 0;
327
Bjorn Helgaas256a4592018-06-08 08:39:45 -0500328 if (!dev->__aer_firmware_first_valid)
329 aer_set_firmware_first(dev);
330 return dev->__aer_firmware_first;
331}
Bjorn Helgaas41cbc9e2018-06-08 08:40:00 -0500332#define PCI_EXP_AER_FLAGS (PCI_EXP_DEVCTL_CERE | PCI_EXP_DEVCTL_NFERE | \
333 PCI_EXP_DEVCTL_FERE | PCI_EXP_DEVCTL_URRE)
Bjorn Helgaas256a4592018-06-08 08:39:45 -0500334
335static bool aer_firmware_first;
336
337/**
338 * aer_acpi_firmware_first - Check if APEI should control AER.
339 */
340bool aer_acpi_firmware_first(void)
341{
342 static bool parsed = false;
343 struct aer_hest_parse_info info = {
344 .pci_dev = NULL, /* Check all PCIe devices */
345 .firmware_first = 0,
346 };
347
Alexandru Gagniuc7af02fc2018-07-03 18:27:43 -0500348 if (pcie_ports_native)
349 return false;
350
Bjorn Helgaas256a4592018-06-08 08:39:45 -0500351 if (!parsed) {
352 apei_hest_parse(aer_hest_parse, &info);
353 aer_firmware_first = info.firmware_first;
354 parsed = true;
355 }
356 return aer_firmware_first;
357}
358#endif
359
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -0500360#define PCI_EXP_AER_FLAGS (PCI_EXP_DEVCTL_CERE | PCI_EXP_DEVCTL_NFERE | \
361 PCI_EXP_DEVCTL_FERE | PCI_EXP_DEVCTL_URRE)
362
363int pci_enable_pcie_error_reporting(struct pci_dev *dev)
364{
365 if (pcie_aer_get_firmware_first(dev))
366 return -EIO;
367
368 if (!dev->aer_cap)
369 return -EIO;
370
371 return pcie_capability_set_word(dev, PCI_EXP_DEVCTL, PCI_EXP_AER_FLAGS);
372}
373EXPORT_SYMBOL_GPL(pci_enable_pcie_error_reporting);
374
375int pci_disable_pcie_error_reporting(struct pci_dev *dev)
376{
377 if (pcie_aer_get_firmware_first(dev))
378 return -EIO;
379
380 return pcie_capability_clear_word(dev, PCI_EXP_DEVCTL,
381 PCI_EXP_AER_FLAGS);
382}
383EXPORT_SYMBOL_GPL(pci_disable_pcie_error_reporting);
384
385int pci_cleanup_aer_uncorrect_error_status(struct pci_dev *dev)
386{
387 int pos;
Oza Pawandeepe7b0b842018-07-19 17:58:05 -0500388 u32 status, sev;
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -0500389
390 pos = dev->aer_cap;
391 if (!pos)
392 return -EIO;
393
Oza Pawandeepe7b0b842018-07-19 17:58:05 -0500394 /* Clear status bits for ERR_NONFATAL errors only */
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -0500395 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
Oza Pawandeepe7b0b842018-07-19 17:58:05 -0500396 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_SEVER, &sev);
397 status &= ~sev;
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -0500398 if (status)
399 pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, status);
400
401 return 0;
402}
403EXPORT_SYMBOL_GPL(pci_cleanup_aer_uncorrect_error_status);
404
Bjorn Helgaas7ab92e82018-07-19 17:55:58 -0500405void pci_aer_clear_fatal_status(struct pci_dev *dev)
406{
407 int pos;
408 u32 status, sev;
409
410 pos = dev->aer_cap;
411 if (!pos)
412 return;
413
414 /* Clear status bits for ERR_FATAL errors only */
415 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
416 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_SEVER, &sev);
417 status &= sev;
418 if (status)
419 pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, status);
420}
421
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -0500422int pci_cleanup_aer_error_status_regs(struct pci_dev *dev)
423{
424 int pos;
425 u32 status;
426 int port_type;
427
428 if (!pci_is_pcie(dev))
429 return -ENODEV;
430
431 pos = dev->aer_cap;
432 if (!pos)
433 return -EIO;
434
435 port_type = pci_pcie_type(dev);
436 if (port_type == PCI_EXP_TYPE_ROOT_PORT) {
437 pci_read_config_dword(dev, pos + PCI_ERR_ROOT_STATUS, &status);
438 pci_write_config_dword(dev, pos + PCI_ERR_ROOT_STATUS, status);
439 }
440
441 pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS, &status);
442 pci_write_config_dword(dev, pos + PCI_ERR_COR_STATUS, status);
443
444 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
445 pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, status);
446
447 return 0;
448}
449
Rajat Jain60ed9822018-06-21 16:48:26 -0700450void pci_aer_init(struct pci_dev *dev)
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -0500451{
452 dev->aer_cap = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
Rajat Jaindb89ccb2018-06-30 15:07:17 -0500453
454 if (dev->aer_cap)
455 dev->aer_stats = kzalloc(sizeof(struct aer_stats), GFP_KERNEL);
456
Rajat Jain60ed9822018-06-21 16:48:26 -0700457 pci_cleanup_aer_error_status_regs(dev);
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -0500458}
459
Rajat Jaindb89ccb2018-06-30 15:07:17 -0500460void pci_aer_exit(struct pci_dev *dev)
461{
462 kfree(dev->aer_stats);
463 dev->aer_stats = NULL;
464}
465
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500466#define AER_AGENT_RECEIVER 0
467#define AER_AGENT_REQUESTER 1
468#define AER_AGENT_COMPLETER 2
469#define AER_AGENT_TRANSMITTER 3
470
471#define AER_AGENT_REQUESTER_MASK(t) ((t == AER_CORRECTABLE) ? \
472 0 : (PCI_ERR_UNC_COMP_TIME|PCI_ERR_UNC_UNSUP))
473#define AER_AGENT_COMPLETER_MASK(t) ((t == AER_CORRECTABLE) ? \
474 0 : PCI_ERR_UNC_COMP_ABORT)
475#define AER_AGENT_TRANSMITTER_MASK(t) ((t == AER_CORRECTABLE) ? \
476 (PCI_ERR_COR_REP_ROLL|PCI_ERR_COR_REP_TIMER) : 0)
477
478#define AER_GET_AGENT(t, e) \
479 ((e & AER_AGENT_COMPLETER_MASK(t)) ? AER_AGENT_COMPLETER : \
480 (e & AER_AGENT_REQUESTER_MASK(t)) ? AER_AGENT_REQUESTER : \
481 (e & AER_AGENT_TRANSMITTER_MASK(t)) ? AER_AGENT_TRANSMITTER : \
482 AER_AGENT_RECEIVER)
483
484#define AER_PHYSICAL_LAYER_ERROR 0
485#define AER_DATA_LINK_LAYER_ERROR 1
486#define AER_TRANSACTION_LAYER_ERROR 2
487
488#define AER_PHYSICAL_LAYER_ERROR_MASK(t) ((t == AER_CORRECTABLE) ? \
489 PCI_ERR_COR_RCVR : 0)
490#define AER_DATA_LINK_LAYER_ERROR_MASK(t) ((t == AER_CORRECTABLE) ? \
491 (PCI_ERR_COR_BAD_TLP| \
492 PCI_ERR_COR_BAD_DLLP| \
493 PCI_ERR_COR_REP_ROLL| \
494 PCI_ERR_COR_REP_TIMER) : PCI_ERR_UNC_DLP)
495
496#define AER_GET_LAYER_ERROR(t, e) \
497 ((e & AER_PHYSICAL_LAYER_ERROR_MASK(t)) ? AER_PHYSICAL_LAYER_ERROR : \
498 (e & AER_DATA_LINK_LAYER_ERROR_MASK(t)) ? AER_DATA_LINK_LAYER_ERROR : \
499 AER_TRANSACTION_LAYER_ERROR)
500
501/*
502 * AER error strings
503 */
504static const char *aer_error_severity_string[] = {
505 "Uncorrected (Non-Fatal)",
506 "Uncorrected (Fatal)",
507 "Corrected"
508};
509
510static const char *aer_error_layer[] = {
511 "Physical Layer",
512 "Data Link Layer",
513 "Transaction Layer"
514};
515
Rajat Jaindb89ccb2018-06-30 15:07:17 -0500516static const char *aer_correctable_error_string[AER_MAX_TYPEOF_COR_ERRS] = {
Tyler Baicarbd237802018-06-26 11:44:15 -0400517 "RxErr", /* Bit Position 0 */
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500518 NULL,
519 NULL,
520 NULL,
521 NULL,
522 NULL,
Tyler Baicarbd237802018-06-26 11:44:15 -0400523 "BadTLP", /* Bit Position 6 */
524 "BadDLLP", /* Bit Position 7 */
525 "Rollover", /* Bit Position 8 */
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500526 NULL,
527 NULL,
528 NULL,
Tyler Baicarbd237802018-06-26 11:44:15 -0400529 "Timeout", /* Bit Position 12 */
530 "NonFatalErr", /* Bit Position 13 */
531 "CorrIntErr", /* Bit Position 14 */
532 "HeaderOF", /* Bit Position 15 */
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500533};
534
Rajat Jaindb89ccb2018-06-30 15:07:17 -0500535static const char *aer_uncorrectable_error_string[AER_MAX_TYPEOF_UNCOR_ERRS] = {
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500536 "Undefined", /* Bit Position 0 */
537 NULL,
538 NULL,
539 NULL,
Tyler Baicarbd237802018-06-26 11:44:15 -0400540 "DLP", /* Bit Position 4 */
541 "SDES", /* Bit Position 5 */
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500542 NULL,
543 NULL,
544 NULL,
545 NULL,
546 NULL,
547 NULL,
Tyler Baicarbd237802018-06-26 11:44:15 -0400548 "TLP", /* Bit Position 12 */
549 "FCP", /* Bit Position 13 */
550 "CmpltTO", /* Bit Position 14 */
551 "CmpltAbrt", /* Bit Position 15 */
552 "UnxCmplt", /* Bit Position 16 */
553 "RxOF", /* Bit Position 17 */
554 "MalfTLP", /* Bit Position 18 */
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500555 "ECRC", /* Bit Position 19 */
Tyler Baicarbd237802018-06-26 11:44:15 -0400556 "UnsupReq", /* Bit Position 20 */
557 "ACSViol", /* Bit Position 21 */
558 "UncorrIntErr", /* Bit Position 22 */
559 "BlockedTLP", /* Bit Position 23 */
560 "AtomicOpBlocked", /* Bit Position 24 */
561 "TLPBlockedErr", /* Bit Position 25 */
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500562};
563
564static const char *aer_agent_string[] = {
565 "Receiver ID",
566 "Requester ID",
567 "Completer ID",
568 "Transmitter ID"
569};
570
Rajat Jain81aa5202018-06-21 16:48:28 -0700571#define aer_stats_dev_attr(name, stats_array, strings_array, \
572 total_string, total_field) \
573 static ssize_t \
574 name##_show(struct device *dev, struct device_attribute *attr, \
575 char *buf) \
576{ \
577 unsigned int i; \
578 char *str = buf; \
579 struct pci_dev *pdev = to_pci_dev(dev); \
580 u64 *stats = pdev->aer_stats->stats_array; \
581 \
582 for (i = 0; i < ARRAY_SIZE(strings_array); i++) { \
583 if (strings_array[i]) \
584 str += sprintf(str, "%s %llu\n", \
585 strings_array[i], stats[i]); \
586 else if (stats[i]) \
587 str += sprintf(str, #stats_array "_bit[%d] %llu\n",\
588 i, stats[i]); \
589 } \
590 str += sprintf(str, "TOTAL_%s %llu\n", total_string, \
591 pdev->aer_stats->total_field); \
592 return str-buf; \
593} \
594static DEVICE_ATTR_RO(name)
595
596aer_stats_dev_attr(aer_dev_correctable, dev_cor_errs,
597 aer_correctable_error_string, "ERR_COR",
598 dev_total_cor_errs);
599aer_stats_dev_attr(aer_dev_fatal, dev_fatal_errs,
600 aer_uncorrectable_error_string, "ERR_FATAL",
601 dev_total_fatal_errs);
602aer_stats_dev_attr(aer_dev_nonfatal, dev_nonfatal_errs,
603 aer_uncorrectable_error_string, "ERR_NONFATAL",
604 dev_total_nonfatal_errs);
605
Rajat Jain12833012018-06-21 16:48:29 -0700606#define aer_stats_rootport_attr(name, field) \
607 static ssize_t \
608 name##_show(struct device *dev, struct device_attribute *attr, \
609 char *buf) \
610{ \
611 struct pci_dev *pdev = to_pci_dev(dev); \
612 return sprintf(buf, "%llu\n", pdev->aer_stats->field); \
613} \
614static DEVICE_ATTR_RO(name)
615
616aer_stats_rootport_attr(aer_rootport_total_err_cor,
617 rootport_total_cor_errs);
618aer_stats_rootport_attr(aer_rootport_total_err_fatal,
619 rootport_total_fatal_errs);
620aer_stats_rootport_attr(aer_rootport_total_err_nonfatal,
621 rootport_total_nonfatal_errs);
622
Rajat Jain81aa5202018-06-21 16:48:28 -0700623static struct attribute *aer_stats_attrs[] __ro_after_init = {
624 &dev_attr_aer_dev_correctable.attr,
625 &dev_attr_aer_dev_fatal.attr,
626 &dev_attr_aer_dev_nonfatal.attr,
Rajat Jain12833012018-06-21 16:48:29 -0700627 &dev_attr_aer_rootport_total_err_cor.attr,
628 &dev_attr_aer_rootport_total_err_fatal.attr,
629 &dev_attr_aer_rootport_total_err_nonfatal.attr,
Rajat Jain81aa5202018-06-21 16:48:28 -0700630 NULL
631};
632
633static umode_t aer_stats_attrs_are_visible(struct kobject *kobj,
634 struct attribute *a, int n)
635{
636 struct device *dev = kobj_to_dev(kobj);
637 struct pci_dev *pdev = to_pci_dev(dev);
638
639 if (!pdev->aer_stats)
640 return 0;
641
Rajat Jain12833012018-06-21 16:48:29 -0700642 if ((a == &dev_attr_aer_rootport_total_err_cor.attr ||
643 a == &dev_attr_aer_rootport_total_err_fatal.attr ||
644 a == &dev_attr_aer_rootport_total_err_nonfatal.attr) &&
645 pci_pcie_type(pdev) != PCI_EXP_TYPE_ROOT_PORT)
646 return 0;
647
Rajat Jain81aa5202018-06-21 16:48:28 -0700648 return a->mode;
649}
650
651const struct attribute_group aer_stats_attr_group = {
652 .attrs = aer_stats_attrs,
653 .is_visible = aer_stats_attrs_are_visible,
654};
655
656static void pci_dev_aer_stats_incr(struct pci_dev *pdev,
657 struct aer_err_info *info)
658{
659 int status, i, max = -1;
660 u64 *counter = NULL;
661 struct aer_stats *aer_stats = pdev->aer_stats;
662
663 if (!aer_stats)
664 return;
665
666 switch (info->severity) {
667 case AER_CORRECTABLE:
668 aer_stats->dev_total_cor_errs++;
669 counter = &aer_stats->dev_cor_errs[0];
670 max = AER_MAX_TYPEOF_COR_ERRS;
671 break;
672 case AER_NONFATAL:
673 aer_stats->dev_total_nonfatal_errs++;
674 counter = &aer_stats->dev_nonfatal_errs[0];
675 max = AER_MAX_TYPEOF_UNCOR_ERRS;
676 break;
677 case AER_FATAL:
678 aer_stats->dev_total_fatal_errs++;
679 counter = &aer_stats->dev_fatal_errs[0];
680 max = AER_MAX_TYPEOF_UNCOR_ERRS;
681 break;
682 }
683
684 status = (info->status & ~info->mask);
685 for (i = 0; i < max; i++)
686 if (status & (1 << i))
687 counter[i]++;
688}
689
Rajat Jain12833012018-06-21 16:48:29 -0700690static void pci_rootport_aer_stats_incr(struct pci_dev *pdev,
691 struct aer_err_source *e_src)
692{
693 struct aer_stats *aer_stats = pdev->aer_stats;
694
695 if (!aer_stats)
696 return;
697
698 if (e_src->status & PCI_ERR_ROOT_COR_RCV)
699 aer_stats->rootport_total_cor_errs++;
700
701 if (e_src->status & PCI_ERR_ROOT_UNCOR_RCV) {
702 if (e_src->status & PCI_ERR_ROOT_FATAL_RCV)
703 aer_stats->rootport_total_fatal_errs++;
704 else
705 aer_stats->rootport_total_nonfatal_errs++;
706 }
707}
708
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500709static void __print_tlp_header(struct pci_dev *dev,
710 struct aer_header_log_regs *t)
711{
712 pci_err(dev, " TLP Header: %08x %08x %08x %08x\n",
713 t->dw0, t->dw1, t->dw2, t->dw3);
714}
715
716static void __aer_print_error(struct pci_dev *dev,
717 struct aer_err_info *info)
718{
719 int i, status;
720 const char *errmsg = NULL;
721 status = (info->status & ~info->mask);
722
723 for (i = 0; i < 32; i++) {
724 if (!(status & (1 << i)))
725 continue;
726
727 if (info->severity == AER_CORRECTABLE)
728 errmsg = i < ARRAY_SIZE(aer_correctable_error_string) ?
729 aer_correctable_error_string[i] : NULL;
730 else
731 errmsg = i < ARRAY_SIZE(aer_uncorrectable_error_string) ?
732 aer_uncorrectable_error_string[i] : NULL;
733
734 if (errmsg)
735 pci_err(dev, " [%2d] %-22s%s\n", i, errmsg,
736 info->first_error == i ? " (First)" : "");
737 else
738 pci_err(dev, " [%2d] Unknown Error Bit%s\n",
739 i, info->first_error == i ? " (First)" : "");
740 }
Rajat Jain81aa5202018-06-21 16:48:28 -0700741 pci_dev_aer_stats_incr(dev, info);
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500742}
743
Keith Busch1e451162018-07-19 16:16:55 -0500744void aer_print_error(struct pci_dev *dev, struct aer_err_info *info)
Bjorn Helgaas0319c9a2018-06-08 08:39:38 -0500745{
746 int layer, agent;
747 int id = ((dev->bus->number << 8) | dev->devfn);
748
749 if (!info->status) {
750 pci_err(dev, "PCIe Bus Error: severity=%s, type=Inaccessible, (Unregistered Agent ID)\n",
751 aer_error_severity_string[info->severity]);
752 goto out;
753 }
754
755 layer = AER_GET_LAYER_ERROR(info->severity, info->status);
756 agent = AER_GET_AGENT(info->severity, info->status);
757
758 pci_err(dev, "PCIe Bus Error: severity=%s, type=%s, (%s)\n",
759 aer_error_severity_string[info->severity],
760 aer_error_layer[layer], aer_agent_string[agent]);
761
762 pci_err(dev, " device [%04x:%04x] error status/mask=%08x/%08x\n",
763 dev->vendor, dev->device,
764 info->status, info->mask);
765
766 __aer_print_error(dev, info);
767
768 if (info->tlp_header_valid)
769 __print_tlp_header(dev, &info->tlp);
770
771out:
772 if (info->id && info->error_dev_num > 1 && info->id == id)
773 pci_err(dev, " Error of this Agent is reported first\n");
774
775 trace_aer_event(dev_name(&dev->dev), (info->status & ~info->mask),
776 info->severity, info->tlp_header_valid, &info->tlp);
777}
778
779static void aer_print_port_info(struct pci_dev *dev, struct aer_err_info *info)
780{
781 u8 bus = info->id >> 8;
782 u8 devfn = info->id & 0xff;
783
784 pci_info(dev, "AER: %s%s error received: %04x:%02x:%02x.%d\n",
785 info->multi_error_valid ? "Multiple " : "",
786 aer_error_severity_string[info->severity],
787 pci_domain_nr(dev->bus), bus, PCI_SLOT(devfn), PCI_FUNC(devfn));
788}
789
790#ifdef CONFIG_ACPI_APEI_PCIEAER
791int cper_severity_to_aer(int cper_severity)
792{
793 switch (cper_severity) {
794 case CPER_SEV_RECOVERABLE:
795 return AER_NONFATAL;
796 case CPER_SEV_FATAL:
797 return AER_FATAL;
798 default:
799 return AER_CORRECTABLE;
800 }
801}
802EXPORT_SYMBOL_GPL(cper_severity_to_aer);
803
804void cper_print_aer(struct pci_dev *dev, int aer_severity,
805 struct aer_capability_regs *aer)
806{
807 int layer, agent, tlp_header_valid = 0;
808 u32 status, mask;
809 struct aer_err_info info;
810
811 if (aer_severity == AER_CORRECTABLE) {
812 status = aer->cor_status;
813 mask = aer->cor_mask;
814 } else {
815 status = aer->uncor_status;
816 mask = aer->uncor_mask;
817 tlp_header_valid = status & AER_LOG_TLP_MASKS;
818 }
819
820 layer = AER_GET_LAYER_ERROR(aer_severity, status);
821 agent = AER_GET_AGENT(aer_severity, status);
822
823 memset(&info, 0, sizeof(info));
824 info.severity = aer_severity;
825 info.status = status;
826 info.mask = mask;
827 info.first_error = PCI_ERR_CAP_FEP(aer->cap_control);
828
829 pci_err(dev, "aer_status: 0x%08x, aer_mask: 0x%08x\n", status, mask);
830 __aer_print_error(dev, &info);
831 pci_err(dev, "aer_layer=%s, aer_agent=%s\n",
832 aer_error_layer[layer], aer_agent_string[agent]);
833
834 if (aer_severity != AER_CORRECTABLE)
835 pci_err(dev, "aer_uncor_severity: 0x%08x\n",
836 aer->uncor_severity);
837
838 if (tlp_header_valid)
839 __print_tlp_header(dev, &aer->header_log);
840
841 trace_aer_event(dev_name(&dev->dev), (status & ~mask),
842 aer_severity, tlp_header_valid, &aer->header_log);
843}
844#endif
845
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -0500846/**
847 * add_error_device - list device to be handled
848 * @e_info: pointer to error info
849 * @dev: pointer to pci_dev to be added
850 */
851static int add_error_device(struct aer_err_info *e_info, struct pci_dev *dev)
852{
853 if (e_info->error_dev_num < AER_MAX_MULTI_ERR_DEVICES) {
854 e_info->dev[e_info->error_dev_num] = dev;
855 e_info->error_dev_num++;
856 return 0;
857 }
858 return -ENOSPC;
859}
860
861/**
862 * is_error_source - check whether the device is source of reported error
863 * @dev: pointer to pci_dev to be checked
864 * @e_info: pointer to reported error info
865 */
866static bool is_error_source(struct pci_dev *dev, struct aer_err_info *e_info)
867{
868 int pos;
869 u32 status, mask;
870 u16 reg16;
871
872 /*
873 * When bus id is equal to 0, it might be a bad id
874 * reported by root port.
875 */
876 if ((PCI_BUS_NUM(e_info->id) != 0) &&
877 !(dev->bus->bus_flags & PCI_BUS_FLAGS_NO_AERSID)) {
878 /* Device ID match? */
879 if (e_info->id == ((dev->bus->number << 8) | dev->devfn))
880 return true;
881
882 /* Continue id comparing if there is no multiple error */
883 if (!e_info->multi_error_valid)
884 return false;
885 }
886
887 /*
888 * When either
889 * 1) bus id is equal to 0. Some ports might lose the bus
890 * id of error source id;
891 * 2) bus flag PCI_BUS_FLAGS_NO_AERSID is set
892 * 3) There are multiple errors and prior ID comparing fails;
893 * We check AER status registers to find possible reporter.
894 */
895 if (atomic_read(&dev->enable_cnt) == 0)
896 return false;
897
898 /* Check if AER is enabled */
899 pcie_capability_read_word(dev, PCI_EXP_DEVCTL, &reg16);
900 if (!(reg16 & PCI_EXP_AER_FLAGS))
901 return false;
902
903 pos = dev->aer_cap;
904 if (!pos)
905 return false;
906
907 /* Check if error is recorded */
908 if (e_info->severity == AER_CORRECTABLE) {
909 pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS, &status);
910 pci_read_config_dword(dev, pos + PCI_ERR_COR_MASK, &mask);
911 } else {
912 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
913 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_MASK, &mask);
914 }
915 if (status & ~mask)
916 return true;
917
918 return false;
919}
920
921static int find_device_iter(struct pci_dev *dev, void *data)
922{
923 struct aer_err_info *e_info = (struct aer_err_info *)data;
924
925 if (is_error_source(dev, e_info)) {
926 /* List this device */
927 if (add_error_device(e_info, dev)) {
928 /* We cannot handle more... Stop iteration */
929 /* TODO: Should print error message here? */
930 return 1;
931 }
932
933 /* If there is only a single error, stop iteration */
934 if (!e_info->multi_error_valid)
935 return 1;
936 }
937 return 0;
938}
939
940/**
941 * find_source_device - search through device hierarchy for source device
942 * @parent: pointer to Root Port pci_dev data structure
943 * @e_info: including detailed error information such like id
944 *
945 * Return true if found.
946 *
947 * Invoked by DPC when error is detected at the Root Port.
948 * Caller of this function must set id, severity, and multi_error_valid of
949 * struct aer_err_info pointed by @e_info properly. This function must fill
950 * e_info->error_dev_num and e_info->dev[], based on the given information.
951 */
952static bool find_source_device(struct pci_dev *parent,
953 struct aer_err_info *e_info)
954{
955 struct pci_dev *dev = parent;
956 int result;
957
958 /* Must reset in this function */
959 e_info->error_dev_num = 0;
960
961 /* Is Root Port an agent that sends error message? */
962 result = find_device_iter(dev, e_info);
963 if (result)
964 return true;
965
966 pci_walk_bus(parent->subordinate, find_device_iter, e_info);
967
968 if (!e_info->error_dev_num) {
969 pci_printk(KERN_DEBUG, parent, "can't find device of ID%04x\n",
970 e_info->id);
971 return false;
972 }
973 return true;
974}
975
976/**
977 * handle_error_source - handle logging error into an event log
978 * @dev: pointer to pci_dev data structure of error source device
979 * @info: comprehensive error information
980 *
981 * Invoked when an error being detected by Root Port.
982 */
983static void handle_error_source(struct pci_dev *dev, struct aer_err_info *info)
984{
985 int pos;
986
987 if (info->severity == AER_CORRECTABLE) {
988 /*
989 * Correctable error does not need software intervention.
990 * No need to go through error recovery process.
991 */
992 pos = dev->aer_cap;
993 if (pos)
994 pci_write_config_dword(dev, pos + PCI_ERR_COR_STATUS,
995 info->status);
996 } else if (info->severity == AER_NONFATAL)
997 pcie_do_nonfatal_recovery(dev);
998 else if (info->severity == AER_FATAL)
999 pcie_do_fatal_recovery(dev, PCIE_PORT_SERVICE_AER);
1000}
1001
1002#ifdef CONFIG_ACPI_APEI_PCIEAER
1003
1004#define AER_RECOVER_RING_ORDER 4
1005#define AER_RECOVER_RING_SIZE (1 << AER_RECOVER_RING_ORDER)
1006
1007struct aer_recover_entry {
1008 u8 bus;
1009 u8 devfn;
1010 u16 domain;
1011 int severity;
1012 struct aer_capability_regs *regs;
1013};
1014
1015static DEFINE_KFIFO(aer_recover_ring, struct aer_recover_entry,
1016 AER_RECOVER_RING_SIZE);
1017
1018static void aer_recover_work_func(struct work_struct *work)
1019{
1020 struct aer_recover_entry entry;
1021 struct pci_dev *pdev;
1022
1023 while (kfifo_get(&aer_recover_ring, &entry)) {
1024 pdev = pci_get_domain_bus_and_slot(entry.domain, entry.bus,
1025 entry.devfn);
1026 if (!pdev) {
1027 pr_err("AER recover: Can not find pci_dev for %04x:%02x:%02x:%x\n",
1028 entry.domain, entry.bus,
1029 PCI_SLOT(entry.devfn), PCI_FUNC(entry.devfn));
1030 continue;
1031 }
1032 cper_print_aer(pdev, entry.severity, entry.regs);
1033 if (entry.severity == AER_NONFATAL)
1034 pcie_do_nonfatal_recovery(pdev);
1035 else if (entry.severity == AER_FATAL)
1036 pcie_do_fatal_recovery(pdev, PCIE_PORT_SERVICE_AER);
1037 pci_dev_put(pdev);
1038 }
1039}
1040
1041/*
1042 * Mutual exclusion for writers of aer_recover_ring, reader side don't
1043 * need lock, because there is only one reader and lock is not needed
1044 * between reader and writer.
1045 */
1046static DEFINE_SPINLOCK(aer_recover_ring_lock);
1047static DECLARE_WORK(aer_recover_work, aer_recover_work_func);
1048
1049void aer_recover_queue(int domain, unsigned int bus, unsigned int devfn,
1050 int severity, struct aer_capability_regs *aer_regs)
1051{
1052 unsigned long flags;
1053 struct aer_recover_entry entry = {
1054 .bus = bus,
1055 .devfn = devfn,
1056 .domain = domain,
1057 .severity = severity,
1058 .regs = aer_regs,
1059 };
1060
1061 spin_lock_irqsave(&aer_recover_ring_lock, flags);
1062 if (kfifo_put(&aer_recover_ring, entry))
1063 schedule_work(&aer_recover_work);
1064 else
1065 pr_err("AER recover: Buffer overflow when recovering AER for %04x:%02x:%02x:%x\n",
1066 domain, bus, PCI_SLOT(devfn), PCI_FUNC(devfn));
1067 spin_unlock_irqrestore(&aer_recover_ring_lock, flags);
1068}
1069EXPORT_SYMBOL_GPL(aer_recover_queue);
1070#endif
1071
1072/**
Keith Busch1e451162018-07-19 16:16:55 -05001073 * aer_get_device_error_info - read error status from dev and store it to info
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001074 * @dev: pointer to the device expected to have a error record
1075 * @info: pointer to structure to store the error record
1076 *
1077 * Return 1 on success, 0 on error.
1078 *
1079 * Note that @info is reused among all error devices. Clear fields properly.
1080 */
Keith Busch1e451162018-07-19 16:16:55 -05001081int aer_get_device_error_info(struct pci_dev *dev, struct aer_err_info *info)
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001082{
1083 int pos, temp;
1084
1085 /* Must reset in this function */
1086 info->status = 0;
1087 info->tlp_header_valid = 0;
1088
1089 pos = dev->aer_cap;
1090
1091 /* The device might not support AER */
1092 if (!pos)
1093 return 0;
1094
1095 if (info->severity == AER_CORRECTABLE) {
1096 pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS,
1097 &info->status);
1098 pci_read_config_dword(dev, pos + PCI_ERR_COR_MASK,
1099 &info->mask);
1100 if (!(info->status & ~info->mask))
1101 return 0;
1102 } else if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE ||
1103 info->severity == AER_NONFATAL) {
1104
1105 /* Link is still healthy for IO reads */
1106 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS,
1107 &info->status);
1108 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_MASK,
1109 &info->mask);
1110 if (!(info->status & ~info->mask))
1111 return 0;
1112
1113 /* Get First Error Pointer */
1114 pci_read_config_dword(dev, pos + PCI_ERR_CAP, &temp);
1115 info->first_error = PCI_ERR_CAP_FEP(temp);
1116
1117 if (info->status & AER_LOG_TLP_MASKS) {
1118 info->tlp_header_valid = 1;
1119 pci_read_config_dword(dev,
1120 pos + PCI_ERR_HEADER_LOG, &info->tlp.dw0);
1121 pci_read_config_dword(dev,
1122 pos + PCI_ERR_HEADER_LOG + 4, &info->tlp.dw1);
1123 pci_read_config_dword(dev,
1124 pos + PCI_ERR_HEADER_LOG + 8, &info->tlp.dw2);
1125 pci_read_config_dword(dev,
1126 pos + PCI_ERR_HEADER_LOG + 12, &info->tlp.dw3);
1127 }
1128 }
1129
1130 return 1;
1131}
1132
1133static inline void aer_process_err_devices(struct aer_err_info *e_info)
1134{
1135 int i;
1136
1137 /* Report all before handle them, not to lost records by reset etc. */
1138 for (i = 0; i < e_info->error_dev_num && e_info->dev[i]; i++) {
Keith Busch1e451162018-07-19 16:16:55 -05001139 if (aer_get_device_error_info(e_info->dev[i], e_info))
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001140 aer_print_error(e_info->dev[i], e_info);
1141 }
1142 for (i = 0; i < e_info->error_dev_num && e_info->dev[i]; i++) {
Keith Busch1e451162018-07-19 16:16:55 -05001143 if (aer_get_device_error_info(e_info->dev[i], e_info))
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001144 handle_error_source(e_info->dev[i], e_info);
1145 }
1146}
1147
1148/**
1149 * aer_isr_one_error - consume an error detected by root port
1150 * @rpc: pointer to the root port which holds an error
1151 * @e_src: pointer to an error source
1152 */
1153static void aer_isr_one_error(struct aer_rpc *rpc,
1154 struct aer_err_source *e_src)
1155{
1156 struct pci_dev *pdev = rpc->rpd;
1157 struct aer_err_info *e_info = &rpc->e_info;
1158
Rajat Jain12833012018-06-21 16:48:29 -07001159 pci_rootport_aer_stats_incr(pdev, e_src);
1160
Bjorn Helgaasfd3362c2018-06-08 08:33:39 -05001161 /*
1162 * There is a possibility that both correctable error and
1163 * uncorrectable error being logged. Report correctable error first.
1164 */
1165 if (e_src->status & PCI_ERR_ROOT_COR_RCV) {
1166 e_info->id = ERR_COR_ID(e_src->id);
1167 e_info->severity = AER_CORRECTABLE;
1168
1169 if (e_src->status & PCI_ERR_ROOT_MULTI_COR_RCV)
1170 e_info->multi_error_valid = 1;
1171 else
1172 e_info->multi_error_valid = 0;
1173 aer_print_port_info(pdev, e_info);
1174
1175 if (find_source_device(pdev, e_info))
1176 aer_process_err_devices(e_info);
1177 }
1178
1179 if (e_src->status & PCI_ERR_ROOT_UNCOR_RCV) {
1180 e_info->id = ERR_UNCOR_ID(e_src->id);
1181
1182 if (e_src->status & PCI_ERR_ROOT_FATAL_RCV)
1183 e_info->severity = AER_FATAL;
1184 else
1185 e_info->severity = AER_NONFATAL;
1186
1187 if (e_src->status & PCI_ERR_ROOT_MULTI_UNCOR_RCV)
1188 e_info->multi_error_valid = 1;
1189 else
1190 e_info->multi_error_valid = 0;
1191
1192 aer_print_port_info(pdev, e_info);
1193
1194 if (find_source_device(pdev, e_info))
1195 aer_process_err_devices(e_info);
1196 }
1197}
1198
1199/**
1200 * get_e_source - retrieve an error source
1201 * @rpc: pointer to the root port which holds an error
1202 * @e_src: pointer to store retrieved error source
1203 *
1204 * Return 1 if an error source is retrieved, otherwise 0.
1205 *
1206 * Invoked by DPC handler to consume an error.
1207 */
1208static int get_e_source(struct aer_rpc *rpc, struct aer_err_source *e_src)
1209{
1210 unsigned long flags;
1211
1212 /* Lock access to Root error producer/consumer index */
1213 spin_lock_irqsave(&rpc->e_lock, flags);
1214 if (rpc->prod_idx == rpc->cons_idx) {
1215 spin_unlock_irqrestore(&rpc->e_lock, flags);
1216 return 0;
1217 }
1218
1219 *e_src = rpc->e_sources[rpc->cons_idx];
1220 rpc->cons_idx++;
1221 if (rpc->cons_idx == AER_ERROR_SOURCES_MAX)
1222 rpc->cons_idx = 0;
1223 spin_unlock_irqrestore(&rpc->e_lock, flags);
1224
1225 return 1;
1226}
1227
1228/**
1229 * aer_isr - consume errors detected by root port
1230 * @work: definition of this work item
1231 *
1232 * Invoked, as DPC, when root port records new detected error
1233 */
1234static void aer_isr(struct work_struct *work)
1235{
1236 struct aer_rpc *rpc = container_of(work, struct aer_rpc, dpc_handler);
1237 struct aer_err_source uninitialized_var(e_src);
1238
1239 mutex_lock(&rpc->rpc_mutex);
1240 while (get_e_source(rpc, &e_src))
1241 aer_isr_one_error(rpc, &e_src);
1242 mutex_unlock(&rpc->rpc_mutex);
1243}
1244
Bjorn Helgaas3c43a642018-06-08 08:31:57 -05001245/**
1246 * aer_irq - Root Port's ISR
1247 * @irq: IRQ assigned to Root Port
1248 * @context: pointer to Root Port data structure
1249 *
1250 * Invoked when Root Port detects AER messages.
1251 */
1252irqreturn_t aer_irq(int irq, void *context)
1253{
1254 unsigned int status, id;
1255 struct pcie_device *pdev = (struct pcie_device *)context;
1256 struct aer_rpc *rpc = get_service_data(pdev);
1257 int next_prod_idx;
1258 unsigned long flags;
1259 int pos;
1260
1261 pos = pdev->port->aer_cap;
1262 /*
1263 * Must lock access to Root Error Status Reg, Root Error ID Reg,
1264 * and Root error producer/consumer index
1265 */
1266 spin_lock_irqsave(&rpc->e_lock, flags);
1267
1268 /* Read error status */
1269 pci_read_config_dword(pdev->port, pos + PCI_ERR_ROOT_STATUS, &status);
1270 if (!(status & (PCI_ERR_ROOT_UNCOR_RCV|PCI_ERR_ROOT_COR_RCV))) {
1271 spin_unlock_irqrestore(&rpc->e_lock, flags);
1272 return IRQ_NONE;
1273 }
1274
1275 /* Read error source and clear error status */
1276 pci_read_config_dword(pdev->port, pos + PCI_ERR_ROOT_ERR_SRC, &id);
1277 pci_write_config_dword(pdev->port, pos + PCI_ERR_ROOT_STATUS, status);
1278
1279 /* Store error source for later DPC handler */
1280 next_prod_idx = rpc->prod_idx + 1;
1281 if (next_prod_idx == AER_ERROR_SOURCES_MAX)
1282 next_prod_idx = 0;
1283 if (next_prod_idx == rpc->cons_idx) {
1284 /*
1285 * Error Storm Condition - possibly the same error occurred.
1286 * Drop the error.
1287 */
1288 spin_unlock_irqrestore(&rpc->e_lock, flags);
1289 return IRQ_HANDLED;
1290 }
1291 rpc->e_sources[rpc->prod_idx].status = status;
1292 rpc->e_sources[rpc->prod_idx].id = id;
1293 rpc->prod_idx = next_prod_idx;
1294 spin_unlock_irqrestore(&rpc->e_lock, flags);
1295
1296 /* Invoke DPC handler */
1297 schedule_work(&rpc->dpc_handler);
1298
1299 return IRQ_HANDLED;
1300}
1301EXPORT_SYMBOL_GPL(aer_irq);
1302
Hidetoshi Seto843f4692010-04-15 13:10:53 +09001303static int set_device_error_reporting(struct pci_dev *dev, void *data)
1304{
1305 bool enable = *((bool *)data);
Yijing Wang62f87c02012-07-24 17:20:03 +08001306 int type = pci_pcie_type(dev);
Hidetoshi Seto843f4692010-04-15 13:10:53 +09001307
Yijing Wang62f87c02012-07-24 17:20:03 +08001308 if ((type == PCI_EXP_TYPE_ROOT_PORT) ||
1309 (type == PCI_EXP_TYPE_UPSTREAM) ||
1310 (type == PCI_EXP_TYPE_DOWNSTREAM)) {
Hidetoshi Seto843f4692010-04-15 13:10:53 +09001311 if (enable)
1312 pci_enable_pcie_error_reporting(dev);
1313 else
1314 pci_disable_pcie_error_reporting(dev);
1315 }
1316
1317 if (enable)
1318 pcie_set_ecrc_checking(dev);
1319
1320 return 0;
1321}
1322
1323/**
1324 * set_downstream_devices_error_reporting - enable/disable the error reporting bits on the root port and its downstream ports.
1325 * @dev: pointer to root port's pci_dev data structure
1326 * @enable: true = enable error reporting, false = disable error reporting.
1327 */
1328static void set_downstream_devices_error_reporting(struct pci_dev *dev,
1329 bool enable)
1330{
1331 set_device_error_reporting(dev, &enable);
1332
1333 if (!dev->subordinate)
1334 return;
1335 pci_walk_bus(dev->subordinate, set_device_error_reporting, &enable);
1336}
1337
1338/**
1339 * aer_enable_rootport - enable Root Port's interrupts when receiving messages
1340 * @rpc: pointer to a Root Port data structure
1341 *
1342 * Invoked when PCIe bus loads AER service driver.
1343 */
1344static void aer_enable_rootport(struct aer_rpc *rpc)
1345{
Keith Busche13d17f2018-04-09 16:04:42 -06001346 struct pci_dev *pdev = rpc->rpd;
Jiang Liu43bd4ee2012-07-24 17:20:11 +08001347 int aer_pos;
Hidetoshi Seto843f4692010-04-15 13:10:53 +09001348 u16 reg16;
1349 u32 reg32;
1350
Hidetoshi Seto843f4692010-04-15 13:10:53 +09001351 /* Clear PCIe Capability's Device Status */
Jiang Liu43bd4ee2012-07-24 17:20:11 +08001352 pcie_capability_read_word(pdev, PCI_EXP_DEVSTA, &reg16);
1353 pcie_capability_write_word(pdev, PCI_EXP_DEVSTA, reg16);
Hidetoshi Seto843f4692010-04-15 13:10:53 +09001354
1355 /* Disable system error generation in response to error messages */
Jiang Liu43bd4ee2012-07-24 17:20:11 +08001356 pcie_capability_clear_word(pdev, PCI_EXP_RTCTL,
1357 SYSTEM_ERROR_INTR_ON_MESG_MASK);
Hidetoshi Seto843f4692010-04-15 13:10:53 +09001358
Keith Busch66b80802016-09-27 16:23:34 -04001359 aer_pos = pdev->aer_cap;
Hidetoshi Seto843f4692010-04-15 13:10:53 +09001360 /* Clear error status */
1361 pci_read_config_dword(pdev, aer_pos + PCI_ERR_ROOT_STATUS, &reg32);
1362 pci_write_config_dword(pdev, aer_pos + PCI_ERR_ROOT_STATUS, reg32);
1363 pci_read_config_dword(pdev, aer_pos + PCI_ERR_COR_STATUS, &reg32);
1364 pci_write_config_dword(pdev, aer_pos + PCI_ERR_COR_STATUS, reg32);
1365 pci_read_config_dword(pdev, aer_pos + PCI_ERR_UNCOR_STATUS, &reg32);
1366 pci_write_config_dword(pdev, aer_pos + PCI_ERR_UNCOR_STATUS, reg32);
1367
1368 /*
1369 * Enable error reporting for the root port device and downstream port
1370 * devices.
1371 */
1372 set_downstream_devices_error_reporting(pdev, true);
1373
1374 /* Enable Root Port's interrupt in response to error messages */
1375 pci_read_config_dword(pdev, aer_pos + PCI_ERR_ROOT_COMMAND, &reg32);
1376 reg32 |= ROOT_PORT_INTR_ON_MESG_MASK;
1377 pci_write_config_dword(pdev, aer_pos + PCI_ERR_ROOT_COMMAND, reg32);
1378}
1379
1380/**
1381 * aer_disable_rootport - disable Root Port's interrupts when receiving messages
1382 * @rpc: pointer to a Root Port data structure
1383 *
1384 * Invoked when PCIe bus unloads AER service driver.
1385 */
1386static void aer_disable_rootport(struct aer_rpc *rpc)
1387{
Keith Busche13d17f2018-04-09 16:04:42 -06001388 struct pci_dev *pdev = rpc->rpd;
Hidetoshi Seto843f4692010-04-15 13:10:53 +09001389 u32 reg32;
1390 int pos;
1391
1392 /*
1393 * Disable error reporting for the root port device and downstream port
1394 * devices.
1395 */
1396 set_downstream_devices_error_reporting(pdev, false);
1397
Keith Busch66b80802016-09-27 16:23:34 -04001398 pos = pdev->aer_cap;
Hidetoshi Seto843f4692010-04-15 13:10:53 +09001399 /* Disable Root's interrupt in response to error messages */
1400 pci_read_config_dword(pdev, pos + PCI_ERR_ROOT_COMMAND, &reg32);
1401 reg32 &= ~ROOT_PORT_INTR_ON_MESG_MASK;
1402 pci_write_config_dword(pdev, pos + PCI_ERR_ROOT_COMMAND, reg32);
1403
1404 /* Clear Root's error status reg */
1405 pci_read_config_dword(pdev, pos + PCI_ERR_ROOT_STATUS, &reg32);
1406 pci_write_config_dword(pdev, pos + PCI_ERR_ROOT_STATUS, reg32);
1407}
1408
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001409/**
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001410 * aer_alloc_rpc - allocate Root Port data structure
1411 * @dev: pointer to the pcie_dev data structure
1412 *
1413 * Invoked when Root Port's AER service is loaded.
Hidetoshi Setof6d37802010-04-15 13:22:11 +09001414 */
Hidetoshi Setoc9a91882009-09-07 17:07:29 +09001415static struct aer_rpc *aer_alloc_rpc(struct pcie_device *dev)
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001416{
1417 struct aer_rpc *rpc;
1418
Hidetoshi Setoc9a91882009-09-07 17:07:29 +09001419 rpc = kzalloc(sizeof(struct aer_rpc), GFP_KERNEL);
1420 if (!rpc)
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001421 return NULL;
1422
Hidetoshi Setof6d37802010-04-15 13:22:11 +09001423 /* Initialize Root lock access, e_lock, to Root Error Status Reg */
Milind Arun Choudharyf5609d72007-07-09 11:55:54 -07001424 spin_lock_init(&rpc->e_lock);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001425
Keith Busche13d17f2018-04-09 16:04:42 -06001426 rpc->rpd = dev->port;
David Howells65f27f32006-11-22 14:55:48 +00001427 INIT_WORK(&rpc->dpc_handler, aer_isr);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001428 mutex_init(&rpc->rpc_mutex);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001429
Stefan Assmann45e829e2009-12-03 06:49:24 -05001430 /* Use PCIe bus function to store rpc into PCIe device */
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001431 set_service_data(dev, rpc);
1432
1433 return rpc;
1434}
1435
1436/**
1437 * aer_remove - clean up resources
1438 * @dev: pointer to the pcie_dev data structure
1439 *
1440 * Invoked when PCI Express bus unloads or AER probe fails.
Hidetoshi Setof6d37802010-04-15 13:22:11 +09001441 */
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001442static void aer_remove(struct pcie_device *dev)
1443{
1444 struct aer_rpc *rpc = get_service_data(dev);
1445
1446 if (rpc) {
1447 /* If register interrupt service, it must be free. */
1448 if (rpc->isr)
1449 free_irq(dev->irq, dev);
1450
Sebastian Andrzej Siewior4ae21822016-01-25 10:08:00 -06001451 flush_work(&rpc->dpc_handler);
Hidetoshi Seto460d2982010-04-15 13:10:03 +09001452 aer_disable_rootport(rpc);
1453 kfree(rpc);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001454 set_service_data(dev, NULL);
1455 }
1456}
1457
1458/**
1459 * aer_probe - initialize resources
1460 * @dev: pointer to the pcie_dev data structure
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001461 *
1462 * Invoked when PCI Express bus loads AER service driver.
Hidetoshi Setof6d37802010-04-15 13:22:11 +09001463 */
Bill Pemberton15856ad2012-11-21 15:35:00 -05001464static int aer_probe(struct pcie_device *dev)
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001465{
1466 int status;
1467 struct aer_rpc *rpc;
Bjorn Helgaas576700b2016-11-21 15:24:25 -06001468 struct device *device = &dev->port->dev;
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001469
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001470 /* Alloc rpc data structure */
Hidetoshi Setoc9a91882009-09-07 17:07:29 +09001471 rpc = aer_alloc_rpc(dev);
1472 if (!rpc) {
Bjorn Helgaas576700b2016-11-21 15:24:25 -06001473 dev_printk(KERN_DEBUG, device, "alloc AER rpc failed\n");
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001474 aer_remove(dev);
1475 return -ENOMEM;
1476 }
1477
1478 /* Request IRQ ISR */
Hidetoshi Setoc9a91882009-09-07 17:07:29 +09001479 status = request_irq(dev->irq, aer_irq, IRQF_SHARED, "aerdrv", dev);
1480 if (status) {
Bjorn Helgaas576700b2016-11-21 15:24:25 -06001481 dev_printk(KERN_DEBUG, device, "request AER IRQ %d failed\n",
1482 dev->irq);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001483 aer_remove(dev);
1484 return status;
1485 }
1486
1487 rpc->isr = 1;
1488
1489 aer_enable_rootport(rpc);
Bjorn Helgaas68a55ae2016-11-21 15:34:02 -06001490 dev_info(device, "AER enabled with IRQ %d\n", dev->irq);
1491 return 0;
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001492}
1493
1494/**
1495 * aer_root_reset - reset link on Root Port
1496 * @dev: pointer to Root Port's pci_dev data structure
1497 *
1498 * Invoked by Port Bus driver when performing link reset at Root Port.
Hidetoshi Setof6d37802010-04-15 13:22:11 +09001499 */
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001500static pci_ers_result_t aer_root_reset(struct pci_dev *dev)
1501{
Hidetoshi Setoc6d34ed2010-04-15 13:09:13 +09001502 u32 reg32;
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001503 int pos;
1504
Keith Busch66b80802016-09-27 16:23:34 -04001505 pos = dev->aer_cap;
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001506
1507 /* Disable Root's interrupt in response to error messages */
Hidetoshi Setoc6d34ed2010-04-15 13:09:13 +09001508 pci_read_config_dword(dev, pos + PCI_ERR_ROOT_COMMAND, &reg32);
1509 reg32 &= ~ROOT_PORT_INTR_ON_MESG_MASK;
1510 pci_write_config_dword(dev, pos + PCI_ERR_ROOT_COMMAND, reg32);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001511
Alex Williamson1b95ce82013-08-08 14:10:20 -06001512 pci_reset_bridge_secondary_bus(dev);
Frederick Lawler7506dc72018-01-18 12:55:24 -06001513 pci_printk(KERN_DEBUG, dev, "Root Port link has been reset\n");
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001514
Hidetoshi Setoc6d34ed2010-04-15 13:09:13 +09001515 /* Clear Root Error Status */
1516 pci_read_config_dword(dev, pos + PCI_ERR_ROOT_STATUS, &reg32);
1517 pci_write_config_dword(dev, pos + PCI_ERR_ROOT_STATUS, reg32);
1518
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001519 /* Enable Root Port's interrupt in response to error messages */
Hidetoshi Setoc6d34ed2010-04-15 13:09:13 +09001520 pci_read_config_dword(dev, pos + PCI_ERR_ROOT_COMMAND, &reg32);
1521 reg32 |= ROOT_PORT_INTR_ON_MESG_MASK;
1522 pci_write_config_dword(dev, pos + PCI_ERR_ROOT_COMMAND, reg32);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001523
1524 return PCI_ERS_RESULT_RECOVERED;
1525}
1526
1527/**
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001528 * aer_error_resume - clean up corresponding error status bits
1529 * @dev: pointer to Root Port's pci_dev data structure
1530 *
1531 * Invoked by Port Bus driver during nonfatal recovery.
Hidetoshi Setof6d37802010-04-15 13:22:11 +09001532 */
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001533static void aer_error_resume(struct pci_dev *dev)
1534{
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001535 u16 reg16;
1536
1537 /* Clean up Root device status */
Jiang Liu43bd4ee2012-07-24 17:20:11 +08001538 pcie_capability_read_word(dev, PCI_EXP_DEVSTA, &reg16);
1539 pcie_capability_write_word(dev, PCI_EXP_DEVSTA, reg16);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001540
Oza Pawandeep5b6c0962018-07-19 17:58:06 -05001541 pci_cleanup_aer_uncorrect_error_status(dev);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001542}
1543
Bjorn Helgaas0054ca82018-06-08 08:31:42 -05001544static struct pcie_port_service_driver aerdriver = {
1545 .name = "aer",
1546 .port_type = PCI_EXP_TYPE_ROOT_PORT,
1547 .service = PCIE_PORT_SERVICE_AER,
1548
1549 .probe = aer_probe,
1550 .remove = aer_remove,
1551 .error_resume = aer_error_resume,
1552 .reset_link = aer_root_reset,
1553};
1554
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001555/**
1556 * aer_service_init - register AER root service driver
1557 *
1558 * Invoked when AER root service driver is loaded.
Hidetoshi Setof6d37802010-04-15 13:22:11 +09001559 */
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001560static int __init aer_service_init(void)
1561{
Rafael J. Wysockib22c3d82010-09-20 18:50:00 +02001562 if (!pci_aer_available() || aer_acpi_firmware_first())
Andi Kleen3e77a3f2009-09-16 22:40:22 +02001563 return -ENXIO;
Sam Ravnborgc1996c22007-02-27 10:22:00 +01001564 return pcie_port_service_register(&aerdriver);
Zhang, Yanmin6c2b3742006-07-31 15:21:33 +08001565}
Paul Gortmaker87563362016-08-24 16:57:46 -04001566device_initcall(aer_service_init);